The Paradoxical Relationship Between Food Insecurity and Obesity

Report Info

Sophia Harris, Sam McFarland, Esal Shakil, and Nika Yermakova

Department of Mathematics, Seattle Pacific University

DAT 4500: Data and Society

Dr. Brian Gill

June 4th, 2024

Abstract

In communities where nutrient-rich foods are a luxury, a paradoxical phenomenon emerges: amidst food insecurity, obesity rates soar. This study delves into the intricate interplay between food insecurity and obesity, aiming to unravel the underlying factors shaping this perplexing relationship. Building upon the insights of the Urban Institute report, our analysis integrates obesity data from the Behavioral Risk Factor Surveillance System with socioeconomic, health, and demographic variables. Through correlation analysis, k-means clustering, and simple moderation analyses, we unveil the nuanced dynamics at play in this relationship. Our findings revealed a moderate positive correlation between food insecurity and obesity. The k-means cluster analysis reflected distinct patterns, shedding light on the varied impacts of socioeconomic factors on food insecurity and obesity. Furthermore, the simple moderation analyses uncover the moderating effects of race, education, and income on this relationship. These insights not only deepen our understanding of the paradox but also pave the way for targeted interventions and policies to break the cycle of food insecurity and obesity in vulnerable communities.

Introduction

Imagine living in a community where grocery stores are scarce, fresh produce is a luxury, and affordable food options are limited to nutrient-poor choices. Now, consider the paradox: despite the lack of food security, obesity rates are soaring. When people think of food insecurity, they imagine a scenario where individuals are lacking access to proper nutrition, so how is this “lacking of food” leading to an excess of weight? This counter-intuitive phenomenon poses a critical question — how can food insecurity coexist with obesity, and what underlying factors contribute to this complex relationship?

Previous research has highlighted an association between obesity rates and food insecurity. However, it has also been identified that this relationship is not straightforward, but a multifaceted issue influenced by numerous factors. While traditional perceptions might suggest that limited access to food would result in decreased calorie intake and subsequently lower rates of obesity, the reality is far more intricate. Factors such as race, education levels, socioeconomic status, and geographic location all play pivotal roles in shaping dietary behaviors and health outcomes within communities experiencing food insecurity. Understanding these factors is important for developing effective interventions and policies to address the relationship between food insecurity and obesity.

The Urban Institute Report

This study was heavily inspired by a report published by the Urban Institute, where they conducted a cluster analysis examining food insecurity and other related factors across counties. The purpose of their study was to disrupt food insecurity by investigating data to identify strategies that tackle the root causes of economic distress and food insecurity. Their report emphasized that solving systemic food insecurity requires more than just providing food; it involves addressing interrelated factors such as housing affordability, employment, financial health, physical health, and geographic barriers.

We decided to extend the Urban Institute’s analysis by including another variable: obesity. We were curious if food insecurity could coexist with obesity and if so, what underlying factors contribute to this relationship. By adding obesity to the cluster analysis, we aimed to deepen the understanding of the inter-dependencies between food insecurity and obesity, building on the insights provided by the Urban Institute’s study.

Food Insecurity

Food insecurity encompasses limited or uncertain access to nutritionally adequate and safe foods, or the inability to acquire acceptable foods in socially acceptable ways (Urban Institute). Frongillo and Bernal note that hunger and malnutrition are common experiences associated with food insecurity. In 2022, the USDA reported that 12.8 percent (17.0 million) of U.S. households experienced food insecurity at some point during the year, reflecting the importance of addressing this issue.

Contrary to popular belief, Franklin et al., emphasize that food insecurity extends beyond mere hunger, focusing also on the quality and adequacy of the diet. Food insecurity is linked to low food expenditures, low fruit and vegetable consumption, and lower-quality diets, all of which can significantly affect health outcomes. Notably, food insecurity without hunger was found to be associated with being overweight among individuals. This suggests that this issue is not just about addressing hunger but also about promoting healthier dietary patterns and weight outcomes (Drewnowski & Specter).

Obesity

The World Health Organization defines obesity as a chronic complex disease characterized by excessive fat deposits that can impair health. Nationally, 41.9 percent of adults have obesity, according to the TFAH. This statistic is alarming because obesity is not merely a cosmetic concern, but it is associated with serious health implications. Individuals with obesity face an increased risk of various chronic diseases, including type 2 diabetes, cardiovascular diseases, hypertension, and certain cancers. Additionally, obesity can adversely affect bone health, reproduction, and overall quality of life, impacting everyday activities such as sleeping or moving (Drewnowski and Specter; Franklin et al.,; World Health Organization; Waidmann et al.,). Furthermore, obesity follows a socioeconomic gradient, disproportionately affecting individuals with limited resources, racial-ethnic minorities, and those living in poverty (Drewnowski and Specter; Burns). This disparity is concerning and follows a similar trajectory reflected in food insecurity.

Current Study

As mentioned previously, we wanted to expand upon the Urban Institute report. We will be doing so through three main objectives. First, we aim to determine if there is a correlation between food insecurity and obesity. Second, we will conduct a k-means cluster analysis that integrates obesity and food insecurity data with socioeconomic, health, and demographic variables to detect emerging patterns. Third, we will perform a set of simple moderation analyses to identify if variables such as race, education, income, and geographic location, moderate the relationship between food insecurity and obesity. Ultimately, we seek to paint a clearer picture of how food insecurity and obesity are intertwined and identify strategies to break this cycle.

Methods and Datasets

Food Insecurity Data

Since we took inspiration from the Urban Institute report, we utilized the data that they posted on their GitHub which included the food insecurity variable, along with other relevant factors. Their study focused on individuals and households in the United States experiencing food insecurity, considering factors such as income, employment, health, housing costs, and demographic characteristics. They used a total of 27 metrics, covering seven broad categories: food insecurity, physical health, housing and transportation costs, income and employment, financial health, demographics, and geography. Most of the metrics were recorded as proportions, with some exceptions, such as median income or credit scores. However, only 19 of the 27 metrics was used to create the county peer groups for their analysis. So, for our cluster analysis, we will be using only those 19 variables.

When it comes to the data collection process, the report utilized various data sets spanning from 2010 to 2018. Because of this, it is difficult to pinpoint what exact data collection procedures were implemented. However, here are the following sources that they used to create their data set:

• Credit Bureau Data (2017): Median credit score, Debt in collections, Medical debt in collections, Average unsecured debt
• US Census Bureau’s Decennial Census and American Community Survey (ACS) (2016 or 2012–16): Median household income, Below 200% of federal poverty level, Unemployment rate, Income inequality, Households with children, Households with seniors (65+), People of color, Black, non-Hispanic, Hispanic, Asian or Pacific Islander, Native American, Another race or multiracial, non-Hispanic, White, non-Hispanic, No college degree, Immigrants, Total population
• County Health Rankings (CHR): Low-birthweight births, People with diabetes, People with disability, No health insurance, Premature deaths per 100,000 people
• Feeding America (2016 and 2017): Food insecure, all people, Food insecure, children, Limited access to healthy food, Households receiving SNAP, Households receiving SNAP or cash assistance
• National Low Income Housing Coalition (NLIHC) (2018): Wage to afford fair market rent
• Center for Neighborhood Technology (CNT) (2017): Transportation costs as percent of income
• Bureau of Labor Statistics (BLS) (2016): Local Area Unemployment Statistics annual-average estimates, Current Population Survey annual-average estimates
• CDC National Vital Statistics System (2010–16): Low-birth weight births data
• CDC Diabetes Interactive Atlas (2014): People with diabetes data
• CDC WONDER (2014–16): Premature deaths per 100,000 people data
• USDA Food Environment Atlas (2015): Limited access to healthy food data

In terms of statistical analysis, the Urban Institute conducted a hierarchical clustering analysis using Ward’s method and a Euclidean distance measure. Their process involved standardizing the 27 metrics across seven broad categories to a comparable scale of 0–100 and normalizing some variables to mitigate the influence of outliers. They analytically down-weighted highly correlated metrics to reduce undue influence and up-weighted key metrics such as food insecurity rates, housing-cost burdens, and rural population share to enhance their impact. Stability analyses were performed to adjust metric weights, ensuring robust group separation and cohesion, as measured by silhouette scores, ultimately defining 10 peer groups.

Obesity Data

While we were interested in somewhat replicating the study conducted by the Urban Institute, we also wanted to include obesity into our analysis. To examine the relationship between food insecurity and obesity, we needed relevant obesity data. We obtained this data from the Behavioral Risk Factor Surveillance System. Since they had data spanning across a multitude of years, we decided to utilize data from 2016, as it aligned best with our food insecurity data.

The data set included our main variable of interest - adult obesity. The CDC defines adult obesity as the percentage of the adult population (ages 18 and older) with a Body Mass Index (BMI) of 30 kg/m² or higher. BMI is calculated using self-reported height and weight with the formula: BMI = weight (kg) / (height (m))². Importantly, the estimates for adult obesity are age-adjusted to the 2000 U.S. standard population to ensure fair comparisons across regions with different age distributions.

The BFRSS employed Random Digit Dialing (RDD) to randomly select participants, generating phone numbers to reach a diverse and representative sample. Surveys were conducted via telephone, targeting both land-line and cellphone users to ensure broad coverage. Data was collected across all states, the District of Columbia, and U.S. territories, focusing on non-institutionalized adults aged 18 and older, living in households with either land-line telephones or cellphones. Trained interviewers conducted the surveys, asking respondents a standardized set of questions, including self-reported height and weight to calculate BMI.

The RDD method guarantees that each adult in the target population has a known, non-zero chance of being selected, making the sample random and representative. With over 400,000 respondents annually since 2011, the large sample size enhances the reliability and generalization of the findings. Data was weighted using Iterative Proportional Fitting (Raking) to adjust for demographic discrepancies and better reflect the population distributions of each state. This ensures the survey sample matches the population in terms of age, sex, race, and other demographics.

Additionally, the BFRSS utilized a multilevel regression and post-stratification (MRP) approach to produce stable estimates, especially for small populations or areas with limited data. Internal and external validation studies confirmed strong consistency between model-based estimates and direct survey estimates. This rigorous methodology further strengthened the reliability of the obesity data used in our analysis.

Data Combination

After obtaining both data sets, we combined them into a single data frame for our statistical analyses and visualizations. Our final data frame included 20 different variables, with food insecurity and obesity as our primary variables of focus. The remaining variables were used as contributing factors of interest. The variables utilized in our study and their corresponding definitions are as follows:

Variable	Definition
Below Poverty	Share of people with incomes below 200 percent of the federal poverty threshold.
Children	Share of households with any children younger than age 18.
College Less	Share of people ages 25 or older with no college degree (includes associate’s degrees, bachelor’s degrees, and graduate degrees).
Credit Score	Median VantageScore credit score ranging from 300 to 850 among those with a credit record.
Debt	Share of people with a credit record who have debt in collections (i.e., severely delinquent debt).
Diabetes	Share of people ages 20 and older with diagnosed diabetes.
Disability	Share of people with at least one disability.
Food Insecure All People	Share of people living in households that experienced food insecurity in the past 12 months.
Food Insecure Children	Share of children living in households that experienced food insecurity in the past 12 months.
Housing Cost Burdened	Share of households (renters and owners) that spend 30 percent or more of their incomes on housing and utility costs.
Low Birthweight	Share of live births with birthweight below 2.5 kilograms (5.5 pounds).
Median Income	Median household income.
No Insurance	Share of people enrolled in neither a private nor public health insurance plan.
Obesity	A chronic complex disease characterized by excessive fat deposits that can impair health.
People Color	Share of people who report being Hispanic or non-Hispanic and one of the following races: Black, Asian or Pacific Islander, American Indian or Alaska Native, another race, or multiracial.
Rural Population	Share of the population living in rural areas.
Seniors	Share of households with a member age 65 or older.
Severely Housing Cost Burdened	Share of households (renters and owners) that spend 50 percent or more of their incomes on housing and utility costs.
Unemployment	Share of people in the labor force who are not employed.
Wage Fair Market Rent	Hourly wage a person working full time, full year needs to make so he or she spends no more than 30 percent of income on rent at HUD’s prevailing fair market rent for a two-bedroom apartment.

Statistical Procedures

All statistical analyses were conducted through RStudio. Descriptive statistics were taken for all variables. The combined data frame underwent a correlation analysis between food insecurity and obesity. We also conducted a k-means cluster analysis, while utilizing the elbow method to determine the optimal number of clusters needed. Lastly, a moderation analysis was conducted for four variables: people of color, less than college education, median income, and rural population. This statistical analysis revealed whether the relationship between food insecurity and obesity is moderated by any of the mentioned variables.

Results

Descriptives

Means, standard deviations, and correlations are presented in the table below.

Descriptive Statistics of All the Variables

	vars	n	mean	sd	median	trimmed	mad	min	max	range	skew	kurtosis	se
State*	1	3135	26.347049	14.084919	25.000000	26.522120	17.791200	1.000000	50.00000	49.00000	-0.0221985	-1.2505910	0.2515565
Abbr*	2	3126	25.279271	13.834255	24.000000	25.366507	17.791200	1.000000	49.00000	48.00000	0.0268003	-1.2736863	0.2474351
County*	3	3135	935.663158	523.502396	935.000000	933.132324	647.896200	1.000000	1867.00000	1866.00000	0.0391392	-1.1191452	9.3497479
Geographic Area*	4	3135	1564.519298	902.932311	1564.000000	1564.598246	1157.910600	1.000000	3128.00000	3127.00000	-0.0004326	-1.1999270	16.1263626
Obesity	5	3135	30.978309	4.451272	31.000000	31.134317	4.447800	11.000000	47.00000	36.00000	-0.3935932	1.0622766	0.0794997
Food Insecure All People	6	3135	13.684880	4.174135	13.100000	13.360064	3.558240	3.700000	36.10000	32.40000	0.9962615	2.1527167	0.0745500
Food Insecure Children	7	3135	20.049506	4.636257	19.900000	19.988641	4.744320	6.200000	40.00000	33.80000	0.1910970	0.2065176	0.0828035
Low Birthweight	8	3035	8.111365	2.041162	7.818930	7.923121	1.743646	2.724796	27.16049	24.43570	1.2188521	3.9469778	0.0370509
Diabetes	9	3135	11.369314	2.516456	11.300000	11.311040	2.520420	3.000000	21.10000	18.10000	0.2414550	0.1419088	0.0449439
Disability	10	3135	15.928637	4.415717	15.421382	15.695488	4.249347	4.655211	36.90263	32.24742	0.5789791	0.5404519	0.0788647
No Insurance	11	3135	11.600325	5.469777	10.963350	11.141509	5.224760	1.420258	52.00365	50.58339	1.1194115	2.8938236	0.0976902
Housing Cost Burdened	12	3135	26.897562	6.482173	26.512236	26.658856	6.137540	6.818182	55.15702	48.33884	0.3774676	0.3749485	0.1157715
Severely Housing Cost Burdened	13	3135	5.738357	3.091249	5.111804	5.384324	2.422741	0.002519	26.74272	26.74020	1.4872237	3.7250005	0.0552097
Wage Fair Market Rent	14	3117	15.413898	4.251324	13.850000	14.593844	1.749468	11.650000	60.02000	48.37000	3.4735123	19.4379991	0.0761475
Median Income	15	3135	48549.160447	13228.505434	46667.000000	47221.600638	10797.775800	18972.000000	134464.00000	115492.00000	1.3334907	3.3737493	236.2609835
Below Poverty	16	3135	37.215814	10.298255	36.859982	37.046919	10.367860	8.279221	74.09811	65.81889	0.1881644	-0.0124096	0.1839267
Unemployment	17	3134	5.221347	1.887911	4.900000	5.041388	1.630860	1.700000	23.60000	21.90000	1.8521355	8.9385816	0.0337234
Credit Score	18	2943	679.227999	36.760922	675.000000	679.979193	37.065000	524.000000	786.00000	262.00000	-0.2019767	0.0520182	0.6776279
Debt	19	2943	33.481681	11.540283	33.246975	33.261758	12.822575	5.102041	68.75000	63.64796	0.1557326	-0.5771490	0.2127264
Children	20	3135	29.704742	5.802479	29.441982	29.523675	4.877874	0.000000	70.06921	70.06921	0.4931028	2.6415390	0.1036322
Seniors	21	3135	31.700275	5.847720	31.495398	31.538808	4.654697	11.804385	78.82235	67.01797	0.5798228	2.7224248	0.1044402
People Color	22	3135	22.922094	19.996752	15.510720	19.872407	15.585224	0.000000	99.23978	99.23978	1.2002044	0.8415898	0.3571418
College less	23	3135	70.392459	10.077813	71.863869	71.341220	9.380131	16.798885	93.10892	76.31004	-0.9559823	1.1598764	0.1799896
Rural Population	24	3134	58.536865	31.473070	59.443673	59.991555	39.458252	0.000000	100.00000	100.00000	-0.1553222	-1.1413279	0.5621984

Correlations

A Pearson product-moment correlation coefficient was computed to assess the relationship between food insecurity and obesity. The correlation coefficient (r = 0.44) indicated that there is a moderate positive linear relationship between food insecurity rates and obesity rates. This means that as food insecurity rates increase, obesity rates tend to increase as well. The t-value of 27.49, along with the very small p-value (< 2.2e-16) reflect that the correlation is statistically significant.

## 
##  Pearson's product-moment correlation
## 
## data:  final_data$food_insecurity_rate and final_data$obesity_rate
## t = 27.49, df = 3133, p-value < 2.2e-16
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
##  0.4121866 0.4686086
## sample estimates:
##      cor 
## 0.440833

Here is a graph to visualize this correlation analysis!

K-means Cluster Analysis

The data set was then subjected to a k-means clustering method to identify distinct groups within the sample based on our twenty variables. Before conducting the analysis we had standardize the data set to make the clustering more accurate. WE also had to remove 405 out of 3135 observations because of missing values within the data set. To determine the optimal number of clusters, we utilized the elbow method. Ultimately, five clusters were determined to be most appropriate for this data set. Here are the characterizations for each of our clusters.

Cluster Characterizations

Cluster	Counties	Food Insecurity	Obesity	Other Characterizations
1	369	Very high	Very high	The poorest, most people of color, least educated.
2	873	Moderately high	Moderately high	Poor, oldest, most disability, very rural, whiter population.
3	305	Very low	Very low	Very rich, urban, educated, and the youngest
4	891	Low	Moderate	Comfortable economically, less urban, very white.
5	475	Moderate	Moderate	Urban, middle class.

Here is a cluster map that was inspired by the Urban Institute, representing a visual relationship of food insecurity and obesity across counties.

Moderation Analysis

A set of simple moderation analyses were conducted to investigate the moderating effects of people of color, less than college education, median income, and rural population on the relationship between food insecurity and obesity.

People of Color

The interaction term (B = 0.0046, p < .001) indicates that the interaction effect increases the impact of food insecurity on obesity as the rates of people of color in a population increase. It is a statistically significant and positive moderator. The R-squared value indicates that 21.87% of obesity is explained by this model.

Populations with higher rates of people of color start with lower obesity rates when there is no food insecurity. However, as food insecurity increases, they experience a steeper rise in obesity. Conversely, whiter populations start off with a higher obesity rate when there is no food insecurity, but the slope is weaker. All the different levels end up crossing at a certain point before diverging in different directions again, which can be confusing.

## 
## Call:
## lm(formula = obesity_rate ~ food_insecurity_rate * people_color, 
##     data = dat)
## 
## Residuals:
##      Min       1Q   Median       3Q      Max 
## -17.0515  -2.1788   0.3736   2.6335  11.1451 
## 
## Coefficients:
##                                    Estimate Std. Error t value Pr(>|t|)    
## (Intercept)                       26.781687   0.417395  64.164  < 2e-16 ***
## food_insecurity_rate               0.344513   0.031921  10.793  < 2e-16 ***
## people_color                      -0.092377   0.010189  -9.067  < 2e-16 ***
## food_insecurity_rate:people_color  0.004595   0.000666   6.900 6.28e-12 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 3.936 on 3131 degrees of freedom
## Multiple R-squared:  0.2187, Adjusted R-squared:  0.218 
## F-statistic: 292.2 on 3 and 3131 DF,  p-value: < 2.2e-16

Less than College Education

The interaction term (B = 0.0052, p < .001) indicates that the interaction effect increases the impact of food insecurity on obesity as less than college education increases. It is a statistically significant and positive moderator. The R-squared value indicates that 41.98% of obesity is explained by this model.

More educated populations start with lower obesity rates when there is no food insecurity and have a weaker slope as food insecurity increases. Less educated populations start with higher obesity rates when there is no food insecurity, and as food insecurity increases, the slope becomes steeper

## 
## Call:
## lm(formula = obesity_rate ~ food_insecurity_rate * college_less, 
##     data = dat)
## 
## Residuals:
##      Min       1Q   Median       3Q      Max 
## -13.3031  -2.1542   0.3013   2.3822  10.4499 
## 
## Coefficients:
##                                    Estimate Std. Error t value Pr(>|t|)    
## (Intercept)                       16.295877   1.359286  11.989  < 2e-16 ***
## food_insecurity_rate              -0.127400   0.104276  -1.222  0.22189    
## college_less                       0.160794   0.018996   8.465  < 2e-16 ***
## food_insecurity_rate:college_less  0.005212   0.001405   3.711  0.00021 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 3.391 on 3131 degrees of freedom
## Multiple R-squared:  0.4203, Adjusted R-squared:  0.4198 
## F-statistic: 756.8 on 3 and 3131 DF,  p-value: < 2.2e-16

Median Income

The interaction term (B = -9.888e-06, p < .001) indicates that the interaction effect decreases the impact of food insecurity on obesity as median income increases. This interaction term is a statistically significant and negative moderator. The model explains 29.26% of the variance in obesity.

Populations with high income show nearly no interaction between food insecurity and obesity. As populations become poorer, the relationship between obesity and food insecurity becomes much stronger.

## 
## Call:
## lm(formula = obesity_rate ~ food_insecurity_rate * median_income, 
##     data = dat)
## 
## Residuals:
##      Min       1Q   Median       3Q      Max 
## -16.1342  -2.0830   0.3492   2.5080  10.9291 
## 
## Coefficients:
##                                      Estimate Std. Error t value Pr(>|t|)    
## (Intercept)                         2.978e+01  6.879e-01  43.291   <2e-16 ***
## food_insecurity_rate                6.072e-01  4.608e-02  13.179   <2e-16 ***
## median_income                      -1.811e-05  1.304e-05  -1.388    0.165    
## food_insecurity_rate:median_income -9.888e-06  1.089e-06  -9.077   <2e-16 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 3.746 on 3131 degrees of freedom
## Multiple R-squared:  0.2926, Adjusted R-squared:  0.292 
## F-statistic: 431.8 on 3 and 3131 DF,  p-value: < 2.2e-16

Rural Population

The interaction term indicates that rural population is not a statistically significant moderator. In other words, changes in rural population levels do not change the interaction between food insecurity and obesity. However, the analysis reveals that rural population has a significant independent effect on obesity (p = 5.48e-08), indicating that changes in rural population levels are associated with an increase in obesity rates.

## 
## Call:
## lm(formula = obesity_rate ~ food_insecurity_rate * rural_population, 
##     data = dat)
## 
## Residuals:
##      Min       1Q   Median       3Q      Max 
## -16.0579  -2.0623   0.3085   2.6629  10.0283 
## 
## Coefficients:
##                                         Estimate Std. Error t value Pr(>|t|)
## (Intercept)                           22.4442893  0.4860741  46.175  < 2e-16
## food_insecurity_rate                   0.4772715  0.0352201  13.551  < 2e-16
## rural_population                       0.0393173  0.0072163   5.448 5.48e-08
## food_insecurity_rate:rural_population -0.0003682  0.0005177  -0.711    0.477
##                                          
## (Intercept)                           ***
## food_insecurity_rate                  ***
## rural_population                      ***
## food_insecurity_rate:rural_population    
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 3.848 on 3130 degrees of freedom
##   (1 observation deleted due to missingness)
## Multiple R-squared:  0.2535, Adjusted R-squared:  0.2528 
## F-statistic: 354.4 on 3 and 3130 DF,  p-value: < 2.2e-16

Discussion

Correlation between food insecurity and obesity

Though it may seem paradoxical at first glance, the correlation between food insecurity and obesity makes sense when we consider a few key factors. The modern food landscape, shaped by industrial processing, floods markets with inexpensive, energy-dense yet nutrient-poor options, disproportionately impacting low-income individuals who often lack access to healthier alternatives (Frongillo & Bernal; Drewnowski & Specter). Economic struggles for food-insecure families worsen the situation by compelling reliance on calorie-dense, budget-friendly options abundant in sugars, fats, and additives, thereby contributing to weight gain and obesity.

Economic factors play a big role here. where households with diminishing incomes prioritize cost-effective foods to maintain energy intake. The affordability of energy-dense options, laden with fats and sugars, leads to overconsumption among food-insecure individuals. This economic necessity drives the preference for cheaper, calorie-dense foods, potentially escalating total energy intake over time. Furthermore, taste preferences for sugar and fat significantly influence dietary choices, with repeated exposure to unhealthy options reinforcing poor eating habits (Drewnowski & Specter).

Moreover, food insecurity often induces chronic stress, prompting maladaptive coping mechanisms like overeating or binge eating. Individuals may seek solace or distraction from economic hardships through food, leading to gradual weight gain. Additionally, the scarcity of food associated with insecurity can strengthen associations between food and reward, intensifying cravings and consumption, perpetuating a cycle of unhealthy behaviors and obesity (Frongillo & Bernal).

Cluster analysis groups interpretations

The K-means cluster analysis reflects interesting patterns.

• Cluster 1: These counties have the most severe levels of food insecurity and obesity. These areas are marked by profound poverty, predominantly inhabited by people of color, and characterized by lower educational attainment. The economic strain significantly restricts access to fresh, nutritious foods, forcing residents to rely on inexpensive, calorie-dense options high in sugars, fats, and additives. This reliance contributes to weight gain and obesity (Frongillo & Bernal). Moreover, racial disparities and systemic inequalities likely worsen food insecurity and obesity in these communities. Studies show that minority populations, particularly Black and Hispanic/Latinx groups, face higher rates of food insecurity and obesity due to systemic racism and inequitable access to resources (Franklin et al.,).

• Cluster 2: These counties have have moderately high levels of food insecurity and obesity. They are characterized by being economically disadvantaged, having an older population, a higher prevalence of disability, and being predominantly rural with a white population. Economic struggles and limited access to healthcare and healthy food options may contribute to these moderate levels of food insecurity and obesity in these predominantly rural counties. Older populations and a higher prevalence of disability may also impact residents’ ability to access and afford healthy food choices, contributing to obesity (Franklin et al.,). Additionally, geographic and economic constraints in rural areas often steer populations towards energy-dense, low-cost foods, increasing susceptibility to obesity (Drewnowski & Specter).

• Cluster 3: These counties have very low levels of food insecurity and obesity. They are characterized by being very rich, urban, educated, and having a younger population. Counties in this cluster benefit from better access to economic resources, education, healthcare, and healthier food options, which contribute to lower levels of food insecurity and obesity. High income levels, urban environments, and a well-educated population likely promote healthier dietary habits and lifestyles, leading to lower obesity rates (Franklin et al.,; Burns).

• Cluster 4: These counties have low levels of food insecurity but moderate levels of obesity. They are characterized by being economically comfortable, less urban, and having a predominantly white population. Despite lower levels of food insecurity, dietary habits, lifestyle choices, and possibly cultural factors may contribute to moderate levels of obesity in these areas. Economic comfort and access to a predominantly white population may influence residents’ food choices and lifestyle behaviors, potentially leading to higher obesity rates (Franklin et al.,).

• Cluster 5: These counties have moderate levels of both food insecurity and obesity. They are characterized by being urban and middle-class. In urban, middle-class areas, access to healthy food options, economic stability, and lifestyle choices likely influence moderate levels of both food insecurity and obesity. Factors such as food availability, affordability, and cultural preferences may impact residents’ dietary habits and contribute to the observed levels of food insecurity and obesity (Franklin et al.,).

Significant moderators

People of color came out to be a significant moderator, suggesting that as the proportion of people of color in a population increases the effect of food insecurity and obesity also increases. This is consistent with existing literature indicating an uneven distribution of food insecurity and obesity across racial and ethnic lines.

• The study by Franklin et al., showed higher rates of food insecurity and obesity among non-Hispanic white and Hispanic groups compared to non-Hispanic blacks.
• Drewnowski and Specter noted elevated rates of obesity and overweight among minority populations, excluding Asian Americans.
• Waidmann et al., further confirms higher obesity rates among Black, Hispanic/Latinx, Native American, Alaska Native, and Pacific Islander communities compared to white and Asian adults.

Additionally,

• the stress associated with economic hardship and social disadvantage, as outlined by Frongillo and Bernal, often leads minority groups to resort to energy-dense comfort foods as coping mechanisms.
• Structural factors, including systemic racism and inequitable resource access, also play part in perpetuating disparities, with links to structural racism impacting healthy food access, healthcare, and other essential resources.

The results also suggest that education level plays a significant role in moderating the relationship between food insecurity and obesity. Specifically, individuals with less than a college education exhibit a stronger interaction effect, indicating that as food insecurity increases, obesity also increases.

• This finding aligns with previous research by Franklin et al.,, who discovered that individuals with lower education levels are not only more likely to experience food insecurity but also have a higher prevalence of obesity.

This relationship can be attributed to several factors elucidated by Burns.

• First, individuals with lower education levels often have limited health literacy, which can impair their ability to make informed dietary decisions.
• Second, education influences dietary behaviors and health awareness, with individuals of higher education levels generally exhibiting healthier behaviors, such as better diet quality and increased physical activity.

The results of our moderation analyses also indicated a decreasing impact of food insecurity on obesity as median income increases. This aligns with the literature, where income levels are consistently highlighted as a crucial variable in discussions surrounding food insecurity and obesity.

• Franklin et al., emphasize the important role of income in moderating this relationship, revealing that low-income households face a heightened risk of food insecurity, consequently contributing to a higher prevalence of obesity within these communities.
• Drewnowski and Specter explain how economic disparities significantly impact diet quality. Their research demonstrates that higher incomes correlate with better diet quality and higher Healthy Eating Index (HEI) scores, while lower incomes are associated with poorer diets characterized by energy-dense, nutrient-poor foods. This economic framework sheds light on how financial constraints shape food choice.
• Furthermore, Burns underscores the socioeconomic gradient of obesity, emphasizing the heightened risk faced by individuals from the lowest income brackets.
• This highlights the challenges encountered by lower-income individuals in accessing healthy foods and adopting healthy lifestyle behaviors.

Despite the analysis indicating that the rural population is not a statistically significant moderator of the interaction between food insecurity and obesity, it does reveal a significant independent effect of the rural population on obesity rates.

• This suggests that changes rural population levels are associated with an increase in obesity rates, independent of their influence on the interaction between food insecurity and obesity.
• This finding is consistent with existing research highlighting the significant impact of geographic conditions, particularly rural environments, on the relationship between food insecurity and obesity.
• Studies like that of Franklin et al., have shown substantial links between food insecurity and elevated obesity rates in rural areas, where limited access to nutritious foods due to geographic and economic constraints often contributes to unhealthy dietary habits.
• Additionally, Trivedi et al., have demonstrated significant disparities in obesity prevalence between rural and urban populations, with rural residents exhibiting higher rates of obesity, emphasizing the unique challenges faced by rural communities in accessing healthy foods and opportunities for physical activity.
• Therefore, while the rural population may not directly moderate the interaction between food insecurity and obesity, its independent effect on obesity rates underscores the complex interplay of geographic and socioeconomic factors in shaping health outcomes.

Limitations

Several limitations need to be acknowledged.

• While Body Mass Index (BMI) is commonly used as an indicator for obesity, it has limitations as an indirect measure of body fat. BMI does not account for factors such as muscle mass or fat distribution, which may result in misclassifying individuals’ obesity status.
• While socioeconomic factors were extensively analyzed, other contextual variables such as cultural norms, food environment, and community resources were not fully employed, leaving potential gaps in understanding this relationship.
• The analysis relies on secondary data sources such as the Urban Institute report and the Behavioral Risk Factor Surveillance System. These datasets may have limitations in terms of accuracy, representativeness, and scope. Additionally, the use of self-reported data, particularly in the BFRSS, introduces the possibility of bias and measurement error.
• There are methodological limitations, where utilizing a multiple moderation analysis over a simple moderation might have been better considering our interest in multiple moderators.
• Due to there being over 500 missing values, this could have influenced the results of the cluster analysis.
• There are other relevant variables that could be included, such as access to healthcare, food assistance programs, community-level interventions, etc.
• The datasets used in the analysis may not fully represent the diversity of populations affected by food insecurity and obesity. For instance, certain demographic groups or geographic regions may be underrepresented or excluded from the analysis, limiting the generalizability of the findings

Conclusion

In conclusion, this study contributes to a deeper understanding of the complex relationship between food insecurity and obesity, highlighting the multifaceted factors that shape dietary behaviors and weight outcomes within communities. By integrating obesity data and conducting rigorous statistical analyses, we reveal nuanced patterns and disparities across socioeconomic and demographic groups, emphasizing the need for targeted interventions and policies to address food insecurity and obesity. Moving forward, efforts to mitigate these intertwined challenges should adopt a holistic approach, addressing structural inequities, promoting access to healthy foods, and empowering communities to make informed choices for better health outcomes.

GitHub Repository

Click here to view our GitHub repository and code!